Transcript How Does Skilled Emigration Affect Developing Countries

Brain drain and development:
an overview
Hillel RAPOPORT
Department of Economics, Bar-Ilan University
and EQUIPPE, University of Lille 2
Presentation for the AFD workshop on “Migration and
Human Capital Development”, Paris, June 30, 2008
(based on joint work with Frederic Docquier)
1
1. Introduction
 By “brain drain” we mean the international mobility
of people with higher (tertiary) education – an elite
in developing countries (about 5% of the workforce)
 Numbers: by 2000 there were 180 million migrants
worldwide, half of them residing in OECD countries.
Of these 90 million migrants, 60 million were aged 25
or more and can be split more or less equally across
education levels (primary, secondary, tertiary).
 Skilled migrants in OECD countries come from
Africa (7%), Asia (35%), Latin America (18%), Eastern
Europe (8%) and from other OECD countries (32%).
2
 Equal split across education categories but different
trends: +70% for skilled migrants in the 1990s
against only +13% for unskilled migrants.
 Three caveats:
- illegal migration: not very serious as we are talking
about STOCKS of SKILLED people
- home/host country education: the definition of a
highly skilled migrant may be either too inclusive or
too exclusive, depending on the outcome of interest
- magnitude (headcounts) v. intensity (emigration
rates): dramatic rise in brain drain magnitudes, but
not necessarily in intensities due to the concomitant
rise in educational attainments in source countries.
3
What are the causes of this increased brain drain?
1. General rise in educational attainments in
developing countries (induces a mechanical rise
even if migration propensities are constant across
education levels)
2. Globalization tends to increase positive selfselection (skilled people agglomerate where human
capital is already abundant, rising skill premium in
some countries – especially the US); however
migration networks are a counter-acting force
3. Selective immigration policies since the 1980s:
point-systems in Australia, Canada et now in the UK,
H1-B Visas in the US, German Green Card, European
Blue Card, French “immigration choisie”.
Should we (they) worry?
4
The answer has long been: YES and NO
 YES: the welfare loss to source countries
goes well beyond the marginal product of the
migrant due to fiscal, technological and
Lucas-type externalities
 NO: there are positive feedback effects
(remittances, return migration, business and
scientific networks).
 How do these positive and negative effects
balance out?
5
 Since the 1970s, the "pessimistic view" has
been dominant.
 Illustration:
"The irony of international migration today is that
many of the people who migrate legally from poor to
richer lands are the very ones that Third World
countries can least afford to lose: the highly educated
and skilled. Since the great majority of these migrants
move on a permanent basis, this perverse brain drain
not only represents a loss of valuable human
resources but could prove to be a serious constraint
on the future economic progress of Third World
nations" (Todaro, 1996: 119).
6
Recent, more optimistic view:
 New theoretical arguments: migration prospects raise
the expected return to education and thus foster
investment in education; when this incentive effect
dominates, the home country can gain.
 New data: Docquier and Marfouk (World Bank, 2005),
Beine, Docquier and Rapoport (WBER2007), Dumont
and Lemaître (OECD WP), Defoort (Population, 2008):
first comparative data sets on emigration rates by
education levels
 New evidence: the first cross-country studies found
evidence of positive effects for some channels (FDI,
technology adoption, human capital formation).
7
Structure of the talk:
 2. Facts: evolution and spatial distribution of
the brain drain
 3. Theory and evidence: on the various
“feedback” channels (remittances, return
migration, business and scientific networks)
 4. Theory and evidence: on the “brain gain”
channel
 5. Conclusions and policy insights
8
2. How big is the brain drain?
Docquier, Lowell and Marfouk (World Bank,
2005): use immigration data from all OECD
countries to compute emigration rates by
education level for about 180 countries in
1990 and 2000. Results:
 Population sizes. Emigration rates decrease
with country size, are 7 times higher in small
countries (10% and 25% for total and skilled
emigration).
 Income levels. Middle income countries have
the highest total and skilled emigration rates,
but the selection bias is highest (12!) in poor
countries.
9
Results for selected countries in 2000:
 Main exporters of brains: UK (1.441 million),
Philippines (1.126), India (1.037), Mexico (0.923),
Germany (0.848), China (0.816).
 Skilled emigration rates >80% in Caribbean and
Central American countries such as Guyana,
Jamaica or Haiti, and >50% in many African
countries. 25 out of the 30 countries with highest
selection biases are in Africa.
 Lowest skilled emigration rates: Turkmenistan
(0.2%), Tadjikistan (0.4%), Saudi Arabia (0.5%),
Bhutan (0.6%) and... the United States (0.5%).
10
Correction for age of entry (Beine, Docquier and
Rapoport (WBER2007)
 Used data on immigrants’ age of entry (as a proxy
for where education was acquired) for 75% of skilled
migrants and estimated the age of entry structure of
the remaining 25% using a gravity model.
 Their results show that controlling for age of entry
has a strong incidence for certain countries, mainly
in Central America.
 However, the rankings based on corrected rates are
roughly similar to the previous ones – suggesting
cross-country comparisons results are robust to the
use of corrected data.
 Examples (+0/+22): Haïti 84/74, Ghana 47/42, Cuba
29/17, Afghanistan 23/11, Mexico 15/10, Poland 13/12
11
Panel data (Defoort, Population2008).
 Skilled emigration rates increased in the 1990s. True




with a longer time-horizon?
Panel data based on the six main immigration
countries, 1975-2000 (one observation every five
years).
Global stability in brain drain intensities as the rise
in educational attainments has pushed selection
biases downwards.
Increases in Central America, East Europe, SubSaharan Africa, South-Central Asia, and decreases in
the Caribbean and North Africa.
Increases in countries such as Brazil, India, Mexico,
or South Africa.
12
Figure 1: Long-run trends in skilled emigration, 1975–2000 (Defoort (2008)
18%
16%
Central America
14%
12%
Sub-Saharan Africa
10%
South-East Asia
8%
Northern Africa
6%
Middle East
South-Cent Asia
4%
South America
Eastern Asia
2%
Eastern Europe
0%
1975
1980
1985
1990
1995
2000
13
Figure 2. Males and females’ brain drain
50,0%
45,0%
40,0%
35,0%
30,0%
M ales
25,0%
Females
20,0%
15,0%
10,0%
5,0%
s
Source: Docquier, Lowell and Marfouk (2007)
M
EN
A
ica
m
er
A
La
tin
Su
b-
Sa
h
Sm
al
l
ar
an
isl
a
A
fr
nd
lo
p
de
ve
Le
as
t
Lo
w
-in
co
ed
m
e
le
id
d
er
-M
Lo
w
pp
U
H
ig
h
er
-M
-in
id
d
co
le
m
e
0,0%
14
3. Theory and evidence:
feedback effects
The “pessimistic” models of the 1970s (e.g.,
Bhagwati and Hamada, JDE1974) were
based on a number of critical assumptions:
(i) Migrants self-select out of the general
population
(ii) No uncertainty on migration opportunities
(iii) Complete disconnection after emigration.
In such conditions, skilled emigration can only
have a negative impact on human capital
formation at home.
15
Waving the above listed assumptions allows
for potentially beneficial effects to kick-in:
 migrants may return after a while
 the education decision may be made in a
context of uncertainty regarding future
migration possibilities
 skilled migrants may send remittances
 skilled migrants may take part in business
and other types of networks that indirectly
benefit the source country.
We discuss these possibilities in the next subsections.
16
3.1. Return and temporary migration
Selective immigration programs are often
intended for temporary migrants; in addition,
temporary migration may be voluntary.
Potential benefits emphasized in recent
theoretical literature:
 Returnees contribute to diffuse the advanced
technology learned abroad (Dos Santos and
Postel-Vinay, JPopE2003)
 Negative selection in return migration that
embodies a brain gain (Stark et al., EL1997)
 Savings, managerial skills (McCormick &
Wahba, EJ2001, Dustmann & Kirchkamp,
JDE2002).
17
Evidence on temporary/return skilled migration
 In general, return migrants are negatively selfselected; return skilled migration is more a
consequence than a trigger of growth.
 Example: proportion of PhDs in Science and
Engineering who return to Taiwan, Korea, China,
India, 1970s/1990s.
 Mixed results from case-studies (e.g., Indian IT
workers: few returnees among engineers
(Saxeenian, 2001), many among high-level workers
(Commander et al., 2004)
 Agrawal, Kapur and McHale (2008): few returnees
among Indian inventors, and perform poorly
 No comparative evidence except for Rosenzweig
(2007)
18
3.2. Remittances
 Remittances are a major source of
disposable income and can relax credit
constraints on human and physical capital
investment.
 Do skilled migrants remit more than
unskilled migrants?
 From the literature we know that the two
main motivations to remit are (familial)
altruism and exchange (generally for
preparing one’s return).
 It is therefore unclear whether the educated
remit more: higher income, but move with
family, on a more permanent basis.
19
Evidence on remittances
Macro data:
 Faini (WBER2007) shows that remittances decrease
with the proportion of skilled migrants. Confirmed
after instrumenting by Nimni, Ozden and Schiff
(2008)
Micro data:
 Kangasniemi et al. (2007): nearly half of Indian
medical doctors in the UK send remittances (16% of
income on average).
 McCormick and Wahba (EJ2001): skill-acquisition
more important for educated migrants than savings
to access to self-employment.
20
3.3. The role of networks
 Business networks: reduces transaction
costs between the host and home countries,
which can favor bilateral trade and FDI
inflows – complementarity v. substitutability
 Scientific networks: allow for the diffusion of
knowledge and favor technology adoption
 Political networks: allow for the diffusion of
democratic values, norms of public behavior,
can lead to better institutions
 Each type of network effect may potentially
compensate the source country for any initial
loss of skills.
21
Evidence on business networks
Trade:
 Bilateral trade: Gould (1994) for the U.S.,
Head and Rees (1998) for Canada: demand
for ethnic goods + networks
 Combes et al. (2005) for intra-regional trade
in France: business/social networks
 Role of ethnic Chinese networks in
international trade (Rauch and Trindade,
2002, Rauch and Casella, 2003) – solve
information problems with trade in
differentiated products differentiated goods.
22
FDI:
 Kugler and Rapoport (2007), US bilateral
data: complementarity between skilled and
services, substitutability between unskilled
and manufacturing FDI
 Confirmed by Javorcik et al. (2007) using IV.
 Kugler and Rapoport (ongoing), bilateral data
for all pairs of countries: same qualitative
results after correction for selection.
 Docquier and Lodigiani (2006): cross-country
comparisons using aggregate emigration by
skill level; positive effect, stronger for
democratic countries and intermediate
corruption levels
23
Evidence on scientific networks
 Meyer (2001) provides lots of anecdotal
evidence on knowledge diffusion and brain
circulation
 Agrawal, Kapur & McHale (2008), use patent
citation data for Indian inventors: co-location
effect larger on average than diaspora effect,
implying a net loss. However the latter is
much stronger for the most cited patents,
which are likely to have the highest value
 Kerr (2008): similar findings for more
diasporas (e.g., Chinese) in more countries
24
4. Brain drain and human capital
formation: gross and net effects
 Starting point: migration is in essence
probabilistic (with internal and external
sources of uncertainty).
 Migration prospects raise the expected
return to education and may thus foster
domestic gross (pre-migration) human
capital formation: brain gain
 If this incentive effect is strong enough to
dominate the brain drain, then there can be a
net gain for the source country
 Theoretical models explore the conditions
required for this to occur -- see Figure 3
25
Figure 3: Probabilistic migration and human
capital formation
P
Pn
0
p*
p’
1
p
26
Micro-level evidence:
 Kangasniemi et al. (SSM, 2007): 30% of Indian MDs
in the UK say emigration prospects affected their
education decisions, respondents say 40% current
medical students in India contemplate emigration.
 Lucas (2004), Philippines: “It is difficult to believe
that these high, privately financed enrolment rates
are not induced by the possibility of emigration.
There are signs that the choice of major field of
study ... responds to shifts in international demands.
Higher education is almost certainly induced to a
significant extent by the potential for emigration''.
 Clemens (2005): dramatic rise in enrollment in
nursing schools in South Africa following recent UK
recognition of foreign education for nurses.
 Batista et al. (2007) for Cape Verde, Gibson,
McKenzie & Stiltman (2008) for Tonga.
27
Macro-level evidence (Beine, Docquier and
Rapoport, EJ2008)
 cross-section of 127 developing countries,
using Docquier and Marfouk (2005).
 First step: testing for the brain gain
hypothesis (i.e., the gross, or ex-ante effect).
 We run a regression of the type:
Δln(Ha,90-00)=a0+a1.ln(Ha,90)+a2.ln(p90)+……+ε
 We find a positive effect of migration on
gross (pre-migration) human capital
formation, with an elasticity of about 5%
(very stable across specifications and
methods – OLS and IV).
 Ongoing work shows findings are robust to
the use of other brain drain and HC
measures, and to alternative functional forms
28
 Second step: net effect (beneficial/detrimental BD)
 Obviously what is important is not how many more
invest in education but how many remain in the
home country
 To compute net effects for each country, we do the
following simulation exercise:
Hcf2000 = Ha2000 - a2.ln(ps90/pu90),
Where Hcf2000 is the counterfactual ex-post HC stock
of the country in 2000, Ha2000 is the observed ex-ante
HC stock (all natives), and ps90/pu90 is the ratio of
skilled to unskilled migration propensities in 1990.
 Hence, Hcf2000 is the net of migration human capital
stock the country would have if skilled workers were
constrained to emigrate with the same probability as
unskilled workers.
29
 Let’s take a numerical example: assume
there are 100 natives in 1990, 20 of them
chose to educated and half of these skilled
leave while out of 80 unskilled workers only
10 have left by 2000.
 Assume emigration rates were also 4 times
higher for the skilled in 1990.
 Then Hp2000 = 10/80 = 12.5%, Ha2000 = 20%, and
Hcf2000 = 0.2 - 0.05*ln(4) = 13%.;
 According to our computations, such a
country has a counterfactual stock which is
higher than its observed stock; it therefore
loses half a percentage point of HC – or a 4%
decrease.
30
 Once we do this simulation for all the
countries in our sample, we find that:
- the losers are characterized by high migration
rates (>20%) and/or high human capital
stocks (also >20%)
- more losers, which tend to lose relatively
more, but winners represent 80% of the
sample population (include China, India,
Indonesia) -- hence the overall absolute net
gain for developing countries.
 Two conclusions: the BD contributes to
increase the number of skilled worldwide and
in the developing world; it is likely to affect
strongly the World Distribution of Income.
31
Figure 4. Brain drain and (net) HC formation
2.0%
Net effect on human capital
1.0%
0.0%
0%
5%
10%
15%
20%
25%
30%
35%
40%
-1.0%
R 2 = 0.6142
-2.0%
-3.0%
Emigration rate of tertiary-educated workers
32
 Beine, Defoort and Docquier (ongoing)
- use Defoort’s (2008) panel data.
- estimate a β-convergence equation for
gross human capital formation, with country
and time fixed effects + interactions between
emigration rates and income group dummies
- results: fixed effects, β (negative), and
interaction between emigration and lowincome status (positive) highly significant
- interpretation: there is convergence, and
the incentive effect is stronger for poor
countries, suggesting credit constraints are
not such a serious issue for higher
education.
33
5. Conclusion
 Micro and macro evidence suggest positive impact
of skilled emigration through business and diaspora
networks (FDI, technology) and some brain gain.
 This is starting to change the way people think about
the brain drain in academic and policy forums
 At a policy level, the fact that the brain drain has
redistributive effects has implications for:
- immigration policy in receiving countries: should
be differentiated across origin countries without
distorting the system too much (legal, moral issues).
- education policy in sending countries : is it still
optimal to subsidize education if people emigrate?
Migration and subsidies can be substitutes
- taxation policy in both: think afresh about the
Bhagwati tax (surplus sharing rather than
compensation, feasibility issues).
34